Financial Contracting in Biotech Strategic

Alliances

David T. Robinson a,∗

aAssistant Professor of Finance and Economics, Graduate School of Business,Columbia University

Toby E. Stuart b

bFred G. Steingraber-A.T. Kearney Professor of Strategy and Organizations,University of Chicago

First Draft

Abstract

We conduct a detailed, micro-level analysis of 126 strategic alliance contracts, all ofwhich were written to sponsor early-stage, genomics-based biotechnology researchat small R&D companies. All contracts prescribe staged investment decisions tocapture the option value associated with the sequential resolution of uncertainty,but the contracts deal with agency problems differently. Among pre-IPO companies,many alliances resemble venture capital contracts: they involve convertible preferredequity and sometimes contain anti-dilution provisions, warrants, and board seats.Contracts contain explicit provisions linking equity participation to subsequent IPOactivity, and contain clauses designed to insulate both parties from multi-taskingproblems. Contrary to the standard assumptions of static contract theory, contractsoften specify provisions that are unobservable or difficult to verify. Equity partici-pation is positively correlated with the ambiguity of the contracting environment.

JEL Classification Codes: G30, G34, G39, M13, O39

∗ Corresponding author. Send correspondence to Uris Hall 414, Columbia Univer-sity, 3022 Broadway, NY, NY 10027, or dtr2001@columbia.edu. We would like tothank Mark Edwards at Recombinant Capital for generously making his data avail-able to us. In addition, we are grateful to George Baker, Bernard Black, Alex Butler,Robert Gibbons, Bengt Holmstrom, Per Stromberg, Andy Wong and participants atthe NBER Strategic Alliance Conference and the Batten Young Scholars conferencefor helpful comments. Sandhya Persad and Weigang Yuan provided expert researchassistance. Any errors are solely our own.

Article submitted to the 2002 Real Options Conference 10 June 2002

1 Introduction

Strategic alliances and joint ventures are an increasingly common vehicle

through which large organizations engage in research and development. For

example, survey evidence from the Pharmaceutical Research and Manufac-

turers of America suggests that roughly 25% of the $26 billion in US-based,

industrially financed, pharmaceutical R&D that occurred in 2000 took place

in over 700 collaborative agreements with outside organizations. This fraction

has tripled since 1991, and has grown twenty-fold since 1981 (National Science

Board, 2000). 1

While the financial and strategic importance of these contracts is widely ac-

knowledged, little is known about their precise structure. In this paper, we

make use of a novel data set to fill this void and conduct a detailed, micro-level

analysis of strategic alliance and joint venture contracts. Our data are propri-

etary contract analyses provided by a major biotechnology industry analysis

firm. The analyses are based primarily on SEC materiality disclosures, but

are supplemented with information culled from industry trade conferences,

scientific meetings, and press releases. Thus, the data are rich in the types of

details that allow us not only to compare alliances to other types of financial

contracts, such as venture capital contracts, but also to compare alliances with

internally conducted R&D projects.

We focus on deals written between large pharmaceutical companies and small,

often start-up research companies in the biotechnology sector. All of the 126

agreements we analyze were written to conduct genomics-based research, and

are distinct from corporate venture capital agreements (see Hellmann (1997)).

1 Moreover, these numbers omit international R&D deals and alliances facilitatedthrough programs such as the Advanced Technology Program. Thus, the true sizeand importance of alliance activity in this sector may be even greater than suggestedhere.

1

Instead, in these deals the pharmaceutical firm is a client, sponsoring a re-

search project that the R&D conducts. A common scientific objective in these

agreements is to identify genetic ‘disease triggers’ that respond to specific

chemical compounds, which can then be developed (with considerable uncer-

tainty) into new drugs.

A deeper understanding of strategic alliance contracts achieves several objec-

tives. First, on a purely practical level, it provides a better understanding of

the organization of industrial research and development. In that regard, this

paper complements the theoretical work of Aghion and Tirole (1994). Given

the hundreds of billions of dollars spent on industrial R&D committed through

federal, academic, and corporate channels, this objective is important in its

own right.

Second, given the inherent ambiguity of the contracting environment, strate-

gic alliances and joint ventures are an ideal empirical setting in which to

explore the predictions of theoretical models on incomplete contracts and op-

timal financial contracting. In this regard, this paper is related to a growing

body of work that explores the predictions of contract theory in various em-

pirical settings (for an interesting survey article, see Chiappori and Salanie

(2000)). The closest papers are Lerner and Merges (1998), which examines

how the allocation of broadly defined control rights to the R&D firm varies

with the availability of outside funding, Kaplan and Stromberg (2000), which

examines term sheets from VC investments in order to examine how these

agreements correspond to various theories of financial contracting (see also

Sahlman (1990)), and Wong (2001), which presents survey evidence on the role

that angel investors play in funding small, nascent firms. As we later show, the

patterns in equity participation that we see in these deals are similar to what

Kaplan and Stromberg (2000) show for venture capital contracts: for pre-IPO

firms, clients often take equity stakes that involve convertible preferred equity

that converts to common stock at IPO. These deals sometimes coincide with

2

board seats, and often involve registration rights and anti-dilution provisions.

Preferred equity is much less common among deals involving publicly traded

R&D firms.

Because the contracting environment is highly uncertain, it is not surprising

that all contracts contain provisions reflecting the option value of abandoning

or waiting to continue. Similar to VC contracts, alliance contracts typically

state at the outset that investments will be staggered and contingent on certain

milestones. In addition to real options, many contracts also contain financial

options: contracts for longer research projects often provide the sponsor with

warrants, and sometimes provide the biotech firm with puts forcing the sponsor

to purchase additional shares in the biotech firm.

However, the alliance contracts in our paper differ from VC contracts in at

least one critical respect: VCs provide funding for firms while alliance partners

sponsor projects inside firms. This means that the contract must provide so-

lutions to multi-tasking problems (Holmstrom and Milgrom, 1991). This gives

rise to a number of important questions. For example, how do firms deal with

multiple, simultaneous collaborations? Do clients distinguish between the suc-

cess of a project and the success of the firm undertaking the project? Our

findings suggest that these questions seem to be at the heart of many alliance

contracts; one of the paper’s main contributions is showing empirically how

multi-task principal-agent problems are solved in practice.

Many contracts contain provisions that seem at odds with the predictions

of contract theory. The contracts we analyze appear rife with clauses and

conditions that are difficult, if not impossible to verify. For example, many

contracts require a certain number of man-hours to be performed in various

activities. Likewise, many contracts state that the pharmaceutical pursue the

development of alliance-developed drugs with the same vigor as their own

drug candidates. Discussions with industry insiders suggest that no explicit

3

mechanisms exist to monitor these considerations. This suggests that implicit

contracts based on relational mechanisms (see, for example, Baker, Gibbons,

and Murphy (2000)) play an important role.

Based on these observations, it seems intuitively clear that a strategic alliance

in biotechnology represents a balance between the added costs of multi-tasking

and asymmetric information against the added gains of project abandonment

value that would not be present were it organized internally. Thus, our final

objective is to shed light on the difference between internally managed research

projects and research carried out under the aegis of an alliance contract. Given

that many contracts specify labor inputs and capital expenditures, many al-

liance projects appear designed to look as if they took place inside the phar-

maceutical firm. In light of this observation, one possible motive for alliances

may be differences in compensation opportunities inside and outside the firm. 2

Another possibility is that contracts are intentionally incomplete, and this in-

completeness enhances the abandonment value associated with conducting the

research with an outside organization as opposed to conducting it internally.

The remainder of the paper is organized as follows. We begin by outlining

the theoretical background. This is contained in Section 2. In Section 3, we

describe the data set we have compiled, while in Section 4 we present details

from the 126 contracts we analyze. Section 6 presents results from regressions

that relate contract characteristics to firm and dyad characteristics. Section

7 concludes by offering future theoretical and empirical research directions

suggested by our findings.

2 See Palia (2002) for evidence that division-level managers in multi-division firmshave lower pay-for-performance sensitivity than managers of similarly sized stand-alone firms.

4

2 Theoretical Motivation

Given that the design of strategic alliances involves elements of organizational

design, optimal contracting, real options, repeated games, and many other

areas of analysis outside economics, an exhaustive review of the relevant lit-

erature is not feasible within the confines of this paper. Instead, this section

address some recent theoretical literature that has predictions for particular

elements of deal structure that we see. We begin by laying out a common

framework for thinking about the transactions we study, then we review re-

sults related to equity participation, control and termination.

2.1 A Common Framework for Alliances

A common element in the deals we analyze is the nature of information and

the timing of actions between the research organization and the pharmaceu-

tical company. All but two of the deals we analyze concern early stage (i.e.

pre-clinical) research. As such, the research company often has specialized

knowledge that the pharmaceutical lacks pertaining to the use of a particular

technology. In the case of genomics-based research, their specialty lies in iden-

tifying gene-based disease triggers that are later screened against a database of

compounds. The pharmaceutical plays a dual role as investor and consumer:

as an investor it uses licensing and royalties, equity participation, or both,

to finance drug development. As a consumer, it takes the R&D firm’s output

and uses it in the further development of a drug. In terms of the nature of the

contracting problem faced by the two parties, a time line common to these

projects is as follows:

Time 0: Would-be alliance partners meet, conduct due diligence, and bargain

over a potential collaboration. If successful, a contract is written at this time.

Time 1: The R&D firm expends effort identifying drug targets. If a target is

5

found, it is passed to the client for further development at time 2.

Time 2: A suitably identified target is transferred to the client firm, which

then integrates this into the discovery and production of some final drug

product.

Time 3: Revenues from the final product occur and are disbursed according

to the agreement.

This paper is of course concerned with the specific nature of the details in

the contract written at time 0, how these details reflect anticipated behavior

at time 1 and 2, and what scope exists for altering or cancelling the contract

after time 0. These considerations are in turn a function of the structure of

information and the nature of the anticipated incentive conflict between the

firms at time 0.

2.2 The Role of Equity

The nature of the activity described above approximates a number of theo-

retical models of incomplete contracts. In particular, the role of the client in

this relationship is a hybrid between the consumer/financier in the Aghion

and Tirole (1994) model and that of the VC in models by Casamatta (2000),

Repullo and Suarez (1999), Cornelli and Yosha (1997).

Aghion and Tirole (1994) model an incomplete contracting situation between

a customer/end user of R&D, and a penniless entrepreneur/researcher who

engages in unverifiable effort to generate an R&D output. In their model, both

agents supply an input (the customer’s input can be thought of as financial

investment) that jointly affects the probability of R&D success, and bargain

over a transfer price in the event that the efforts result in success. Their results

echo the standard prescriptions from Grossman and Hart (1986) and Hart

and Moore (1990), namely that when contracts are incomplete, ownership

6

should be tilted towards the agent whose marginal impact on the value of

the project is highest. This is perhaps the most compelling explanation for

why these contracts occur between two firms, and not within a firm, where

ultimate ownership would not rest with the researcher. 3 In addition, their

theory also provides for ‘shop rights,’ which in our context coincide with the

R&D firm’s right to own (for the purposes of later development) rights to

certain compounds that are not selected by the client.

But a number of features of the Aghion and Tirole (1994) model seem at

odds with the data we present here. Most importantly, in their model equity

ownership is irrelevant, since it has no effect on the real transfer price of the

R&D output: any combination of equity and a license fee can be mimicked by

a lower equity stake and a license fee adjusted upward to reflect the difference

in the license fee not internalized through ownership. As they explicitly point

out, however, a host of questions concerning ownership and management of

research processes and mitigating problems with spill-overs may give rise to

the need for equity.

Explicit motives for equity participation arise in Repullo and Suarez (1999)

and Casamatta (2000), where a double-sided moral hazard problem arising

from the financiers dual role as investor and adviser makes the use of eq-

uity desirable. Casamatta (2000) shows that in general, when the financier

also provides a complementary input (advisory services in her model) equity

is necessary to provide the financier with appropriate incentives. Her model

develops a prediction that common equity should occur when the financier’s

investment is low, but convertible preferred equity should occur when the

financier’s investment is high.

3 Aghion and Bolton (1992) provides another motive for why these activities occurin alliances as opposed to within the firm–namely, when the private benefits to theR&D are co-monotonic with the social value of the project (the sum of private andmonetary benefits to both parties), ownership by the R&D is optimal.

7

3 Data

The data we use for this study come from a database (www.rdna.com) as-

sembled by Recombinant Capital, a biotechnology industry analysis firm that

provides access to a wide range of contract-related information based on data

culled from public filings, news releases, and presentations at industry confer-

ences. Recombinant Capital not only tracks inter-firm collaborations in human

medicine, but also agreements involving universities, and collaborations in re-

lated fields such as agricultural technology and veterinary medicine. In order

to remove one potential source of contractual variation, we focus exclusively

on genomics deals initiated between a drug and a biotech, or between two

biotech firms prior to 1998. Broadly speaking, genomics involves using ad-

vances in biology and genetics to understand disease processes at the cellular

level. Functional genomics specifically entails locating genes that contribute

to disease in affected cells.

Genomics companies work at the first stage of the contemporary drug de-

velopment process: they identify “drug targets”—enzymes or receptors that

trigger or block biochemical processes within a cell. The biological role of

these targets in disease initiation or progression is then “validated,” a process

which entails proving that a DNA, RNA, or protein molecule directly partici-

pates in a disease process and is therefore a suitable target for development of

a new therapeutic compound. Validated targets are then “screened” against

(typically hundreds of thousands) molecules, with the aim of pinpointing com-

pounds that trigger or block the processes precipitated by the focal targets.

In all of the alliances we examine in detail, the biotech partner identifies and

validates targets, which are then developed in collaboration with the client.

In some of the partnerships, the biotech partner will also screen compounds

against targets, and thus transfer lead development compounds to the client.

Although biotech firms continue to expand downstream in the drug develop-

8

ment chain, the client in the partnership typically conducts the subsequent

steps in the drug development process, including animal testing, clinical tri-

als, large-scale manufacturing, and sales and marketing. Roughly speaking,

one can think of these alliances as vertical transactions in which there is an

upstream / downstream division of effort between the biotech firm and the

client in the deal.

The sample was created by searching the rdna.com database on the key-

words “Combinatorial,” “Gene Expression,” “Gene Sequencing,” “Pharma-

cogenomics,” “Proteomics,” “Screening,” and “Transcription Factors.” In ad-

dition, we restricted attention to deals that were already analyzed by Recom-

binant Capital, meaning that Recombinant Capital employees had synthe-

sized the SEC filings and news announcements into a common document for-

mat. Using these screens yielded 218 deals, some of which seem inappropriate

for the present analysis, given that they are primarily licensing agreements

for already-existing products (for instance, granting access to a proprietary

database). For these deals, it is not clear what the intra-firm alternative to

the alliance is, and what organizational implications the alliance may have on

either counterparty. Since our objective is to understand complex inter-firm re-

lationships, not genomics per se, we exclude 66 such deals based on subjective

evaluation, leaving a total of 152 deals. Twenty-six further deals were omitted

because they were later-stage deals, leaving 126 deals in our final sample.

One shortcoming to our approach is that it is ultimately based on publicly

available information, and thus many confidential terms are hidden from us.

At the same time, we can conduct tests based on market reactions and subse-

quent performance that are not normally available in detailed studies of other

types of contracts. This allows us to build on previous results that highlight

the role of equity as a mechanism for allocating control (Pisano, 1989; Robin-

son and Stuart, 2000; Boone, 2001). Whereas many papers simply assume that

equity stakes confer control rights, citing the incidence of board seats granted

9

in conjunction with larger equity stakes, our analysis provides details which

sharpen this intuition. The propensity of equity deals to involve detailed pro-

hibitions on amassing shares in the open market subsequent to the alliance

suggests that equity confers control even when it is not accompanied by board

seats, voting rights, or other similar measures.

In part, our sampling strategy reflects the limitations associated with work-

ing with data derived from publicly available sources. Due to SEC disclosure

provisions, many dollar amounts and percentages are omitted from our docu-

ments. 4 On the other hand, since many of the firms in our sample are publicly

traded, we have access to information (financials, stock price reactions) that

is often lacking in the study of venture activity.

Table 1 shows the sample characteristics over time. While alliances appear

every year from 1990 to 1998, roughly 75% of the alliances take place after

1994. The client takes an equity position in the target in 81 deals. Upfront

payments appear in 35 of 126 deals. At the same time, the mean size of the

deal (a number which includes upfront payments as well as contingent pay-

ments that may not occur, but nevertheless approximates the total potential

commercial value to the R&D of the project) trends upward too. If we were

to extend this sample forward past 1998, we would see genomics-related al-

liance activity explode–a fact which coincides in part with the push to map

the human genome.

Table 2 provides details on the biotech firms involved in these deals. Of the 62

biotech firms represented, 46 were privately held at least once in our sample.

Thirty-two firms remain private throughout our sample, and sixteen are public

when they enter our sample, leaving fourteen that enter our sample as private

firms but appear in our sample again as a publicly held firm. An additional ten

4 Lerner and Merges (1998) overcome this problem by simply measuring the numberof different types of control mechanisms allocated to the biotech firm.

10

firms IPO during the period, but for these firms we do not record their alliance

activity after they become public. The number of deals involving publicly

traded trends upward through the sample, reflecting the stock of prior IPOs

in biotechnology.

Comparing Tables 1 and 2 illustrates the relative size of the alliance com-

pared to the size of the R&D firm. The total estimated value of all contingent

payments that represented in the alliance (Mean Size in Table 1) is roughly

half the size of the recorded value of the R&D firm as reported in Table 2.

Comparing the mean equity stake in Table 1 to the mean firm size in Table

2, we see that the equity stakes comprise roughly 5% of R&D firm value.

4 A Closer Look at Contract Characteristics

In this section, we examine the features of these contracts by focusing not only

on specific examples that illuminate the difficulties inherent in the alliance,

but also by presenting tabulations of key contract characteristics.

For clarity, we focus our three specific contracts in detail, and then use these

deals as a platform for discussing the board summary statistics presented in

the tables. These are the Biogen/Curagen deal signed in 10/1997; the Bristol

Myers Squibb/Cadus deal signed in 7/1994; and the Millennium/Bayer deal

signed on 9/1998.

4.1 R&D Staging

Every contract in our sample involves the use of staged financing that coin-

cides with research milestones. The BMS/Cadus deal specified a three-year

research period with an option to extend, while the other two specified five-

year research periods. The BMS/Cadus deal also specified that up to $4 million

11

per year in research funding would be provided for Cadus, and according to its

1996 annual report, it had received $10.4 million in funding from BMS during

the first two years of the agreement.

For 68 deals it is possible to determine the expected length of the collaboration

based on the initial contract. This is presented in Table 4. The mean and

median are both approximately four years, but the maximum length is ten

years, and this occurs in three deals. This length reflects the expected time

that both parties anticipate the alliance will last at the inception of the project,

as reflected by the contract. Of the remaining 7 deals, only the deal between

Ortho Biotech and Cell Therapeutics did not specify an alliance length. The

other 6 were either confidential or not available.

Another important part of the terms of R&D is the provision for labor alloca-

tion. The Biogen/Curagen and Millennium/Bayer deals both specify a certain

number of full-time equivalents (FTEs) to be devoted to the project, while

the BMS/Cadus deal does not. Generally, the number of FTEs is confidential

in our data, however its very presence is interesting given the potential diffi-

culty in verifying that labor is actually being supplied. That these contracts

include what might be regarded as unverifiable actions suggests that theories

of implicit contracts along the lines of Baker, Gibbons, and Murphy (1998)

may help describe why alliances are successful.

Indeed, Table 1 indicates that of the 126 deals included in our analysis, 57 spec-

ify the number of full-time equivalents (FTEs) devoted to the research project.

Is this labor input verifiable? Perhaps not, but twelve alliance contracts go fur-

ther and state that a specific grade or education level be used–for example,

this might state that the personnel be appropriately qualified in biochemistry

or biology, or that they hold Ph.D.s. In addition, sixteen contracts specify

that certain, named personnel be employed strictly on a particular project,

that they not be allowed to work on other projects, and that if they should no

12

longer be employed, that the deal should be renegotiated. This indicates that

the contracting parties seem aware of the inalienability of the human capital

involved in the research process, as suggested in Hart and Moore (1994).

The source of incentive conflict in many financing theories is private bene-

fits; in models of internal capital markets such as Stein (1997), Stein (2000),

and others, managers are motivated by non-appropriable private benefits. One

potential source of private benefits in the deals we analyze are publications

arising from discoveries. These scientific publications seem like an important

of compensation for scientists and are inherently tied to project outcomes.

In fact, many scientists at biotech firms are prolific contributors to the sci-

entific literature. Evidence from Stern (1999) indicates that these scientists

forgo substantial wage income for the ability to publish in scientific journals.

Interestingly, the terms of the R&D section of the contracts provide guidelines

for publishing academic articles based on scientific discoveries related to the

alliance. While many deals place strict prohibitions on this activity, others

permit it given appropriate permissions have been obtained. In the Millen-

nium/Bayer deal, notice must be

“given 60 days prior to submission to other party. If the other Party informs

such Party that its proposed publication could be expected to have a mate-

rial adverse effect on any Patent Rights or Know-How of such other Party,

then such Party shall delay such proposed publication sufficiently long to

permit the timely preparation and first filing of patent application(s) on

the information involved. Millennium shall not permit a publication that

includes information relating to a Bayer Development Candidate without

the prior approval of Bayer.”

The BMS/Cadus and Biogen/Curagen deals simply specify written prior con-

sent.

13

4.2 Collaboration Management

Collaboration management is an area in which surprisingly little variation

across contracts exists. Equal representation on management committees is

the norm, and majority or unanimity is almost always required in order to act

on a decision.

The Millennium/Bayer, Curagen/Biogen, and BMS/Cadus deals contain very

similar provisions for collaboration management. In all three cases, the deal

specifies equal representation from the client and R&D on the project-level

committees. In the BMS/Cadus deal, Cadus was responsible for appointing

the project coordinator, while the others simply state equal numbers at each

level of decision-making. Decisions in the Millennium/Bayer deal are made on

the basis of majority opinion, whereas the other two deals require unanimous

opinion, and the Millennium/Bayer deal specifies project-level committees as

well as a joint steering committee. The distinction, then, is whether a single

member of an opposing firm can block a decision, or whether all members of

the opposing firm must act in concert to block a decision of the committee.

A far greater degree of oversight appears in these deals relative to what one

would expect in venture capital agreements. Moreover, this oversight is not

part of an initial control stake maintained by the VC which fades as the project

matures; if anything, given the vertical nature of the relationship between the

R&D firm and the client, the value of control may in fact increase as the

project matures.

4.3 Equity and other Financing

A key element to the deals we analyze is equity cross-ownership. As Allen

and Phillips (2000) show, equity ownership in strategic alliances is common

14

across a wide range of business activities. The deals we describe fit the pattern

laid out in Allen and Phillips (2000), in which the larger firm takes an equity

position in the smaller firm as part of the funding of the collaboration it

sponsors. In this section, we focus not only on the presence of equity, but also

with the contractual arrangements that surround the equity stake.

The Millennium/Bayer deal included just over $96M in equity for Millennium

Pharmaceuticals. With this came the right for a Bayer representative to at-

tend Millennium board meetings semi-annually. The terms of their agreement

specified that Bayer would pay 115% of the maximum of (i) the average Mil-

lennium stock price between March 1, 1998 and August 31, 1998, and (ii) the

stock price on 21 September, 1998. In fact, the historical average was higher.

The deal was signed on 22 September, 1998. Millennium’s stock price dy-

namics around that date were as follows: on Friday, 18 September, it traded

at $14.31/share. The Monday, 21 September, price jumped to $16.50, and

the price on the 22nd continued upward to $17.25. This high price continued

throughout the year.

One of the interesting features of this transaction was the manner in which

contingencies were written into the equity agreement. Since privately negoti-

ated equity placements such as these cannot generally be liquidated through

public sale for two years and without prior registration, the assignment of de-

mand and piggyback registration rights is potentially important. 5 The con-

tract provided Bayer with two demand registrations and unlimited piggyback

registrations. In addition to these rights, Bayer was prohibited from transfer-

ring or selling more than 2.5 million shares in any one year—this prohibition

was erased if Millennium entered into a merger agreement. Bayer was entitled

to maintain its pro rata share ownership in Millennium, but only if Bayer had

5 Demand rights allow the holder to force the other firm to register its stock for sale;piggyback rights allow the holder to include its shares in any registration initiatedby the other firm.

15

not sold more than one million shares over the life of the agreement. Not only

was Bayer restricted in its ability to sell Millennium stock; Bayer was also

prohibited from increasing its share ownership in Millennium for three years

after the signing of the contract.

BMS provided Cadus with $20 million in three separate equity transactions.

The first two transactions involved $12.5 million (in July, 1994) and $5 million

(in September, 1995) of Class B convertible preferred stock, purchased at a

share price of $3.50 and $4, respectively. The second equity purchase occurred

as a result of Cadus achieving a research milestone. Finally, at the IPO of

Cadus in July, 1996, BMS converted its B shares into 1.607 million common

shares, and purchased an additional $2.5 million worth of common shares at

$7/share.

The Biogen/Curagen deal is unusual in that it combines equity and debt.

Biogen purchased $5 million of common equity in Curagen at its IPO price

of $11.50/share. The terms of the purchase were such that if Curagen did not

IPO within 18 months of the deal date, Biogen had no obligation to purchase

further stock–effectively it acted as a large shareholder in the IPO. It also

provided a $10 million loan facility. The loan was repayable in cash or Curagen

common stock (at current market prices) at the sole discretion of Curagen.

More evidence on the role of equity can be found in Table 3. The striking

feature of Table 3 is the distinction between equity positions in pre- and post-

IPO firms. Among privately held firms, equity is typically structured as con-

vertible preferred securities. Convertible preferred equity holdings are roughly

twice the size of common equity positions taken in pre-IPO firms. Conversion

to common equity typically takes place at IPO, and in 15 of the 54 deals, the

contract contained provisions requiring the pharmaceutical client to purchase

more equity in the public offering. Relatively few deals (7 of 54) involve debt.

16

4.4 Licensing and Termination

A key feature in almost all biotech strategic alliances is the licensing agreement

that supports the exchange of revenues between the companies once a drug

candidate has been identified. In terms of the time line at the beginning of

this section, the licensing agreement takes effect at time 2, once the R&D

output has been transferred upstream, and specifies the behavior of the client

at time 3. One of the interesting features of strategic alliances in biotechnology,

as suggested by Table 1 [in future draft] and as also shown in Robinson and

Stuart (2000), is that the licensing agreement is written at the inception of the

contract, before the object of the license exists. The alternative to this, which

would perhaps be more natural from the point of view of contract theory,

would be to postpone the licensing agreement until a discovery materialized.

It seems noteworthy that contracts typically do not state that the parties

agree to determine licensing terms at a later date.

Each of the three deals described here provides an exclusive license to the

client for any compounds that are identified as suitable candidates. (Identi-

fying candidates for the client, after all, is the primary objective of the al-

liance.) In the Millennium/Bayer deal, Bayer received an exclusive license

with respect to selected targets, but only a non-exclusive license with respect

to targets that were returned from the selection process. Unfortunately, the

royalty rates associated with these licenses are confidential, but we do ob-

serve variation in the manner that revenues are divided between the firms: the

Millennium/Bayer and BMS/Cadus deals confer worldwide licensing rights to

clients for all disease categories, while the Biogen/Curagen licensing revenues

are split according to disease category.

Termination rights are a central part of the theories of Bolton and Scharfstein

(1990) and Hart and Moore (1998), in which the outside financier’s ability

to shut down the entrepreneur’s project at some intermediate stage (before

17

unobservable cash flows arrive) provides the entrepreneur with incentives not

to enjoy too many private benefits.

There are essentially two aspects of termination provisions. One concerns who

is allowed to terminate the deal, and under what circumstances. The second

concerns what happens to the existing intellectual property after the termi-

nation.

Regarding the first point, the three alliances we scrutinize differ in how they al-

locate termination rights. The BMS/Cadus deal allocates termination rights to

both the client and the R&D equally–each may terminate by material breach

only. In the Biogen/Curagen deal, Curagen’s right to terminate is limited to

uncured material breach, but Biogen has substantially more rights. Biogen too

has the right to terminate for uncured material breach or bankruptcy, but also

may terminate any time after the second anniversary with six months’ written

notice. In this event, the contract states that any such early termination shall

not affect any license agreement.

The deals also differ in terms of what happens to existing alliance resources

after the termination. The Biogen/Curagen deal effectively states that all re-

sources that are not part of an ongoing license shall be returned to the original

creator of the resource–all client proprietary material shall be returned to the

client or destroyed, likewise with all R&D material. The other two deals make

no such provision, except to state that existing licenses outlive any termina-

tion. Finally, the three deals coincide in their treatment of transferring tech-

nology: each deal prohibits one firm from transferring technology to a third

party without written consent of the counterparty, except in the event of a

sale or merger.

These findings suggest that termination rights do more than simply provide

incentives. Termination rights seem to play an important role in the manner in

which intellectual property and other resources are appropriated at the project

18

level, and kept from being implemented in other projects that one firm have

in operation.

Table 4 provides evidence on how alliances end. It shows that the most com-

mon outcome is for ownership to revert back to the R&D after a project

has been terminated, provided that the contract did not end due to R&D

breach. That such a small fraction of alliances involve sharing arrangements

supports Aghion and Bolton (1992), who show that typically co-ownership

is sub-optimal relative to contingent ownership, since the former exacerbates

holdup problems.

5 What Is Missing From These Contracts?

Up to this point, we have focused on what is present in strategic alliance con-

tracts: how financing terms and collaboration management are used together

to manage investment projects that are carried out in separate firms. In this

section, we ask what might be present in these contracts but instead is miss-

ing. In order to do that, we compare alliance contracts with other types of

financial contracts that have received recent empirical attention.

5.1 Strategic Alliances and VC Funding

Because these contracts often represent the most important means of financ-

ing for small biotechnology firms, venture capital contracts provide a natural

point of comparison. Alliance agreements look similar to venture capital agree-

ments in a number of respects: they involve staged capital infusions based on

performance milestones that mirror the financing rounds in VC deals; they fre-

quently involve the use of equity, and in particular, convertible preferred stock;

and monitoring occurs frequently through board representation. Like Kaplan

19

and Stromberg (2000), we find that these contracts separately allocate control

rights, cash-flow rights, liquidation rights, and board rights.

However, the contracts we analyze here specify VC-like controls far less fre-

quently than do standard VC contracts. Kaplan and Stromberg (2000) report

that nearly all VC deals will guarantee at least one board seat; in their sam-

ple, a significant fraction actually obtain a board majority. In our sample,

less than 20% of the deals with pre-IPO firms involve board seats, and in no

circumstance did the deal specify a majority of seats.

Yet these contracts differ from venture capital deals in important ways. The

fundamental distinction is that VCs fund the development of firms, whereas

alliance agreements fund the development of projects inside firms. This means

that in alliance agreements, significant resources go into delineating the ac-

ceptable use of resources in non-project related activities in a way that is not

present in VC deals. Prototypical deals between VCs and entrepreneurs typ-

ically involve the VC maintaining cash flow rights but ceding control as the

entrepreneur’s business matures. Not so with alliance contracts. Due to the

vertical nature of the relationship between the R&D firm and the pharmaceu-

tical organization, trading off control and residual income over the life of the

project is less common. For example, collaboration management is typically

handled by a project team comprising members in equal numbers from both

the biotech and the pharmaceutical. Decision making is often by unanimous

vote, but given the ubiquity of equal representation, even majority voting allow

either party to block decisions they view as inappropriate. This differs from

what we see in venture capital, where VCs frequently sit on boards (Lerner,

1995) and have voting rights (Sahlman, 1990; Kaplan and Stromberg, 2000),

but do not have day-to-day, operational decision-making rights.

20

6 Regression Results

In this section we present multivariate results relating key contract character-

istics to other terms of the contract and the contracting environment. Because

all the contract characteristics are determined endogenously, it is difficult to

establish a direction of causation; instead, these regression results should be

interpreted as conditional correlations, expanding the results presented in ta-

ble 5.

6.1 Ambiguity and Contract Length

Table 6 presents results from Tobit regressions on the length of the contract–

i.e., the number of words required to describe the terms and conditions of the

alliance–on a number of contract and dyad characteristics. Since we are not

using the physical contracts directly, it is important to describe the dependent

variable more carefully.

The dependent variable is constructed by measuring the length (in bytes)

of the contract analysis provided by Recombinant Capital. Recall that the

contract analyses are presented in a standard format. Every contract in our

sample is represented by a computer file with exactly 63 lines, the length of

which vary with the complexity of the description of that line item. Therefore,

if a contract contains no provisions for the use of equity, then a number of lines

of the contract analysis are simply left blank. Similarly, if the nature of the

alliance activity is difficult to describe and requires specifying a great many

contingencies, then the contract analysis will contain lengthy descriptions in

the fields devoted to the description of research. Thus, by measuring the byte-

length of the contract, we capture the complexity of the contract in a simple,

yet robust manner. Measuring the contract according to its size on a computer

instead of counting the words in the contract has the added feature that

21

lengthy, technical descriptions are ascribed more length than a description

with the same word count using simpler words. Also, using contract analyses

rather than raw contracts means that we our measures are not confounded by

legal terms of art that are lengthy but communicate relatively simple, well-

known ideas.

Given the relatively narrow scope of the alliance activity in questions, it may

come as a surprise that there is considerable variation in contract length. The

mean is 20 kilobytes, and the median is 18 kilobytes, but the fifth and 95th

percentiles span from 10 to 35 kb. We take contract length to be a proxy

for the difficulty in specifying all the possible contingencies that can arise in

the course of conducting the activity in question; likewise, contract length

measures the complexity required to describe each possible contingent action.

What explains contract length? We find larger equity stakes coincide with

longer contracts. This supports the idea that equity is used as a control device.

(Note that we are controlling for whether the deal involves equity, to control

for the possibility that deals with equity are longer simply because they must

specify the terms of the equity transaction.) Also, we find that when larger

upfront payments are made to the R&D, contracts are longer. Interestingly,

longer contracts coincide with more frequent meetings of the research oversight

committees that monitor the progress of the collaboration.

6.2 Funding Provisions

6.2.1 Upfront Payments

Based on the pairwise correlations presented in table 5, we see that the amount

of upfront funding is positively (and significantly) correlated with the size of

the project and its stage at inception, positively correlated with the value

of the biotech firm, and negatively correlated with the length of the project.

22

This suggests that the level of upfront funding is an outcome of the relative

bargaining power of the two parties involved and the level of ambiguity of the

research activity.

Table 7 presents regressions of upfront payment, in millions of US dollars, on

various deal characteristics. Regardless of the controls included in the regres-

sion, we see that contracts with later stage projects specify larger amounts of

upfront funding for the biotechnology firm. While it is certainly the case that

the two Phase I stage projects had large upfront payments, excluding them

does not change this finding. A simple comparison of mean funding amounts by

stage shows that discovery stage projects receive upfront payments at roughly

the same frequency as lead-molecule or pre-clinical stage deals (31% for discov-

ery stage versus 35% for the latter types), but the upfront funding is roughly

half the size ($1.18 million versus $2.02 million).

Biotech centrality in Table 7 is a weighted measure of the number of past

transactions that the firm has engaged in. It is designed to capture two phe-

nomena: how many alliances a firm has formed in the past, and how influential

are the prior alliance partners. The basic idea behind the measure is that hav-

ing formed an alliance with an influential partner is more valuable in terms of

reputation building than having formed an alliance with a peripheral partner–

one who has done relatively few deals, and with more peripheral firms. In order

to capture this intuition, each past transaction is weighted by the centrality

of that transaction’s counterparty, which is in turn calculated in an analogous

manner (see Robinson and Stuart (2000) for mathematical details). Thus, al-

liances with more central firms are more highly weighted, and alliances formed

with peripheral firms are less highly weighted. That biotech centrality is posi-

tively correlated with the magnitude of the upfront payment supports the idea

that the size of the upfront payment is an outcome of a bargaining process in

which biotechs with more bargaining power get more favorable deal terms.

23

Table 7 also shows that, conditional on firm age and centrality, upfront pay-

ments are a great deal smaller in the presence of equity participation. This

suggests that equity and upfront payments are substitutes: that one role of

equity participation is to mask an upfront cash transfer to the biotech by

disguising it as an equity participation. 6

7 Conclusion

This paper examines the details of 126 strategic alliance agreements. In each

of the 126 cases, the collaboration centers on using one firm’s expertise in

genomics to identify and discover drug candidates that will later become part

of the other firm’s product pipeline.

Because these deals provide a major source of funding to small, nascent (often

pre-IPO) firms, part of our analysis compares these contracts with venture

capital deals. Like venture capital deals, the projects often involve staged

capital infusions triggered by successful completion of milestones, the use of

convertible, preferred equity and debt, and the allocation of monitoring rights

through board membership.

But the comparison with venture capital does not fully explain the complex-

ities of the inter-firm collaborative agreements we study. An important com-

ponent of the contract is delineating the boundaries of the project of interest,

and keeping its resources and revenues separate from the rest of the firm re-

sponsible for its execution. Thus, we commonly see project-level operational

decisions made in teams of equal numbers from both firms. Contracts also

clearly specify rights pertaining to the use of intellectual property that arises.

This research highlight the tradeoffs (in terms of agency costs) that agents

6 This characterization is commonly asserted by corporate development officersresponsible for many of these deals.

24

face when they try to implement organizational structures that are designed to

capture abandonment value and other real options value inherent in sequential

investment decisions. One important area for future work lay in developing

a sharper understanding of this tradeoff. How do organizational mechanisms

designed to capture option value in projects balance the benefits abandonment

value, sequential investment, and the option to delay against agency costs and

other costs introduced in multi-tasking environments?

This suggests that an important direction for theoretical progress on alliances

and ventures lies in understanding the mechanisms behind verifying resource

allocation between projects at the research firm. Organizing an investment

project as a strategic alliance with an outside firm rather than as an internally

managed project seems to optimize the tradeoff between gaining abandonment

value through the alliance and facing greater multi-tasking and asymmetric in-

formation problems by placing organizational boundaries between the project

and its funding source. Gaining a better understanding of such inter-firm col-

laboration is likely to increase our understanding of internal capital markets

and decision-making inside firms, as well as shed light on the issues relating

to the determinants of the boundaries of the firm.

25

References

Aghion, P., and P. Bolton (1992): “An incomplete contracts approach tofinancial contracting,” The Review of Economic Studies, 59(3).

Aghion, P., and J. Tirole (1994): “The Management of Innovation,” Quar-terly Journal of Economics, 109(4), 1185–1209.

Allen, J. W., and G. M. Phillips (2000): “Corporate Equity Owner-ship, Strategic Alliances, and Product Market Relationships,” Journal ofFinance, 55(2).

Baker, G., R. Gibbons, and K. Murphy (2000): “Relational Contractsand Strategic Alliances,” Working Paper, Harvard Business School.

Baker, G., R. Gibbons, and K. J. Murphy (1998): “Implicit Contractsand the Theory of the Firm,” Working paper, NBER.

Bolton, P., and D. Scharfstein (1990): “A Theory of Predation Based onAgency Problems in Financial Contracting,” American Economic Review,80(1), 94–106.

Bonacich, P. (1987): “Power and centrality: A family of measures,” Amer-ican Journal of Sociology, 92, 1170–1182.

Boone, A. (2001): “The Interaction of Ownership, Governance, and ProductMarkets: Evidence from Equity Carveouts,” Working Paper, Penn StateUniversity.

Casamatta, C. (2000): “Financing and Advising: Optimal Financial Con-tracts with Venture Capitalists,” CEPR Discussion Paper.

Chiappori, P.-A., and B. Salanie (2000): “Testing Contract Theory:A Survey of Some Recent Work,” Invited lecture, World Congress of theEconometric Society, Seattle, Washington.

Cornelli, F., and O. Yosha (1997): “Stage Financing and the Role ofConvertible Debt,” CEPR Discussion Paper 1735.

Grossman, S., and O. Hart (1986): “The costs and benefits of ownership:a theory of vertical and lateral integration,” Journal of Political Economy,94(4), 691–719.

Hart, O., and J. Moore (1990): “Property Rights and the Nature of theFirm,” Journal of Political Economy, 98.

(1994): “A Theory of Debt Based on the Inalienability of HumanCapital,” Quarterly Journal of Economics, 109(4), 841–879.

(1998): “Default and Renegotiation: A Dynamic Model of Debt,”Quarterly Journal of Economics, 113(1), 1–41.

Hellmann, T. (1997): “A Theory of Corporate Venture Investing,” WorkingPaper No. 1452, Stanford University.

Holmstrom, B., and P. Milgrom (1991): “Multitask Principal-AgentAnalyses: Incentive Contracts, Asset Ownership, and Job Design,” Jour-nal of Law, Economics, and Organization, 7, 24–51.

Kaplan, S. N., and P. Stromberg (2000): “Contracting Theory MeetsReality: An Empirical Investigation of Venture Capital Contracts,” Workingpaper, University of Chicago.

26

Lerner, J. (1995): “Venture Capitalists and the Oversight of Private Firms,”Journal of Finance, 50(1), 301–318.

Lerner, J., and R. P. Merges (1998): “The control of strategic alliances:an empirical analysis of biotechnology collaborations,” Journal of IndustrialEconomics.

National Science Board (2000): “Science and Engineering Indicators –2000,” Discussion Paper NSB-00-1, National Science Foundation, Arlington,VA.

Palia, D. (2002): “Corporate Governance and the Diversification Discount:Evidence from Panel Data,” Working Paper, Rutgers University.

Pisano, G. (1989): “Using equity participation to support exchange: evi-dence from the biotechnology industry,” Journal of Law, Economics, andOrganization, 5(1).

Repullo, R., and J. Suarez (1999): “Venture Capital Finance: A SecurityDesign Approach,” Working Paper, CEMFI.

Robinson, D. T., and T. E. Stuart (1999): “Experience, Control, andValue Creation in Biotech Strategic Alliances,” Working paper in progress,University of Chicago.

(2000): “Just How Incomplete Are Incomplete Markets? Evidencefrom Biotech Strategic Alliances,” Working Paper, University of Chicago.

Sahlman, W. A. (1990): “The Structure and Governance of Venture-CapitalOrganizations,” Journal of Financial Economics, 27, 473–521.

Stein, J. C. (1997): “Internal Capital Markets and the Competition for Cor-porate Resources,” Journal of Finance, 52(1), 111–133.

(2000): “Information Production and Capital Allocation: Decentral-ized vs. Hierarchical Firms,” Working Paper.

Stern, S. (1999): “Do Scientists Pay to be Scientists?,” NBER WorkingPaper w7410.

Wong, A. (2001): “Angel Investors,” Dissertation, University of Chicago,Chicago, IL 60637.

27

Tab

le1

Con

trac

tC

hara

cter

isti

csov

erT

ime

This

table

des

crib

esth

etim

e-se

ries

ofallia

nce

announce

men

tsin

our

sam

ple

,alo

ng

with

oth

erkey

sam

ple

chara

cter

istics

.Tota

ldea

lsis

the

num

ber

ofannounce

men

tsin

yea

rt,

while

mea

nsi

zeis

the

mea

nofto

talple

dged

fundin

gas

reco

rded

by

Rec

om

bin

ant

Capital(t

his

incl

udes

contingen

tfu

ndin

g).

Expec

ted

Len

gth

isth

enum

ber

ofyea

rsth

at

the

collabora

tion

isex

pec

ted

tola

st.E

quity

dea

lsare

the

num

ber

inw

hic

hth

ecl

ient

purc

hase

deq

uity

inth

eta

rget

,w

hile

size

isth

em

ean

size

inm

illions

ofdollars

.‘%

Share

’is

the

size

ofth

eeq

uity

stake

expre

ssed

as

afr

act

ion

ofth

eto

talfirm

valu

e,as

mea

sure

dby

Rec

om

bin

ant

Capit

al.

The

‘Up-F

ront

Fee

?’co

lum

ns

reco

rdth

enum

ber

each

yea

rth

at

involv

edupfr

ont

cash

paym

ents

.FT

Es

spec

ified

reco

rds

whet

her

the

contr

act

spec

ifies

the

num

ber

of

man-h

ours

tobe

spen

ton

rese

arc

h;‘P

ay

Gra

de

Nam

ed’co

unts

whet

her

thes

em

an-h

ours

are

tobe

per

form

edby

scie

ntist

sw

ith

spec

ific

train

ing

or

back

gro

und

(e.g

.,P

hD

bio

chem

ists

).‘P

erso

ns

Nam

ed’co

unts

whet

her

aco

ntr

act

isco

ntingen

ton

the

per

form

ance

ofa

spec

ific,

nam

edin

div

idualem

plo

yed

at

the

R&

Dfirm

.Scr

eenin

gdea

lsare

those

inw

hic

hth

eco

ntr

act

spec

ifies

that

the

R&

Dsc

reen

sta

rget

sagain

sta

data

base

ofco

mpounds.

Tot

alM

ean

Expec

ted

Equity

Dea

ls:

Upfr

ont

Em

plo

ym

ent:

Scr

eenin

g

Yea

rD

eals

Siz

eLen

gth

Cou

nt

Siz

e%

Shar

eFee

?FT

Es

Pay

-Gra

de

Per

sons

Dea

l

1990

129

.50

5.00

15.

0048

%0

00

10

1991

816

.88

3.25

43.

008.

5%3

10

04

1992

1114

.17

4.23

73.

189.

4%4

43

27

1993

929

.21

3.19

63.

345.

5%4

41

16

1994

1448

.85

3.64

117.

267.

9%1

71

17

1995

2745

.45

3.63

184.

876.

4%7

133

512

1996

1541

.61

3.58

103.

706.

0%6

91

17

1997

2954

.49

3.98

196.

337.

5%8

122

514

1998

1260

.53

4.11

51.

581.

5%2

71

07

Ove

rall

126

37.8

53.

8581

4.25

7.1%

3557

1216

64

28

Tab

le2

Cha

ract

eris

tics

ofSa

mpl

eR

&D

Fir

ms

over

Tim

e

This

table

des

crib

esth

ech

ara

cter

istics

ofR

&D

firm

sin

our

sam

ple

.Tota

lFir

ms

isth

enum

ber

ofdis

tinct

firm

sin

our

sam

ple

ina

giv

enyea

r.Fir

mValu

eis

the

valu

eofth

eR

&D

firm

,as

reco

rded

by

Rec

om

bin

ant

Capital.

This

figure

isbase

don

the

last

reco

rded

valu

ation

for

pri

vate

firm

s,and

isoth

erw

ise

the

mark

etvalu

eofeq

uity.

‘Then

Publicl

yTra

ded

’re

cord

sth

enum

ber

ofR

&D

firm

sth

at

wer

epublicl

ytr

aded

at

the

tim

eofth

eannounce

men

t,w

hile

‘Late

rP

ublicl

yTra

ded

’re

cord

sth

enum

ber

ofnon-p

ublicl

ytr

aded

firm

sth

at

late

rw

ent

public

aft

erth

eti

me

ofth

eallia

nce

.‘P

erce

nt

ven

ture

’re

cord

sth

e(m

ean)

fract

ion

of

R&

Dfirm

’ssh

are

souts

tandin

gow

ned

by

VC

inte

rest

s.P

rior

dea

lsm

easu

res

all

corp

ora

tetr

ansa

ctio

ns

reco

rded

by

the

Rec

om

bin

ant

Capit

al

fundin

gdata

base

–th

isin

cludes

pri

vate

equity

inves

tmen

ts,ven

ture

fundin

g,and

oth

erso

urc

esoffu

ndin

g,but

excl

udes

IPO

sand

opti

on

or

warr

ant

exer

cise

s.P

rior

allia

nce

sm

easu

res

pre

vio

us

allia

nce

act

ivity

inth

ree

ways.

‘Last

5Yea

rs’m

easu

res

tota

lallia

nce

act

ivity

(with

both

oth

erbio

tech

nolo

gy

firm

sas

wel

las

pharm

ace

utica

ls),

‘Over

all’m

easu

res

all

reco

rded

allia

nce

act

ivity

for

the

firm

,w

hile

‘Equity’act

ivity

mea

sure

sall

pri

or

equity

allia

nce

act

ivity.

Tot

alFir

mFir

mP

ublic

Pri

vate

Per

cent

Pri

orP

rior

Allia

nce

s:

Fir

ms

Val

ue

Age

At

Dea

lth

enP

ub?

VC

Dea

lsLas

t5

Yrs

.O

vera

llE

quity

1990

110

.42.

580

114

%6.

001.

00.

00.

0

1991

778

.71.

612

457

%4.

143.

02.

61.

1

1992

1013

0.6

3.19

22

43%

5.30

3.8

4.9

2.5

1993

878

.23.

272

559

%7.

113.

94.

41.

1

1994

1284

.63.

945

749

%7.

094.

87.

42.

8

1995

1995

.44.

637

1254

%7.

914.

76.

72.

4

1996

1210

1.8

4.91

67

50%

7.50

5.8

7.7

1.7

1997

2317

1.9

4.67

158

41%

8.70

7.3

11.0

3.2

1998

1029

0.2

5.54

75

32%

10.5

00.

00.

00.

0

102

115.

83.

8246

5144

%7.

143.

85.

01.

6

29

Table 3The Use of Equity in Alliance Agreements

This table summarizes equity participation in biotech strategic alliances for our sample of 127 firms. Panel A containsdata for all 127 firms; Panel B only summarizes information for firms that were not publicly traded at the time ofthe alliance, while Panel C summarizes information for the complement. Total deals is the number of deals containingequity, common equity, or preferred equity. Convertible denotes the subset of preferred equity transactions that couldbe identified as convertible equity. Mean and median amount are in millions and refer to the size of the equity stake.Mean Fraction expresses the equity stake as a fraction of the total R&D firm value, which is based on the firm’svaluation in its last venture round. Board seats is the number of equity deals in which a board seat is granted tothe client firm as part of the deal (in all but one case, a single seat is given). IPO tie-in refers to whether the initialequity stake is part of a planned IPO of the R&D firm. This includes situations in which the alliance coincides withthe R&D’s IPO, and when the alliance calls for the client to increase its equity stake at the time of the IPO. Loantie-in refers to deals in which the equity stake is tied to the repayment of a loan provided to the R&D by the client.

Total Mean Median Mean Board IPO Loan

Deals Amount Amount Fraction Seats Tie-Ins Tie-Ins

Panel A: All Firms, N=126

Total 82 7.39 5.4 11 11 15 11

Common 40 8 5 6.5 5 6 8

Preferred 42 6.85 6 14.95 6 9 3

Convertible 32 7.22 6.25 14.6 4

Panel B: Pre-IPO Firms, N=72

Total 54 5.93 5 12.05 9 15 7

Common 14 4.35 4 7.8 3 6 5

Preferred 40 6.45 6 13.7 6 9 2

Convertible 30 6.7 6.2 13.06 4

Panel C: Post-IPO Firms, N=55

Total 28 10.43 8.25 9 2 4

Common 26 10.43 8.25 6.4 2 3

Preferred 2 14.5 14.5 32.3 0 1

Convertible 2 14.5 14.5 32.3 0

30

Table 4How Alliances End: Termination Provisions and Ownership Reversion

This table describes termination provisions and ownership reversion upon termination for a sample of 126 strategic alliancecontracts. Ownership reverts to R&D includes all situations in which ownership reverts to the R&D, including cases in whichownership reverts with exceptions. (54 of 75 contracts stipulate that ownership reverts to R&D with no further language.)

Both/ Only Only

Termination Provision Either R&D Client Examples

Uncured Breach Only 0 4 0

Uncured Breach 100 0 0 Either party may terminate by breach, bybankruptcy, by mutual agreement, or if theother party is acquired by any third party

If Change in Control 11 0 38

At Will 4 0 16

At Will after Certain Date 8 0 51 . . .may be terminated by [Client] any timeafter the 3rd anniversary of signing; . . .

If Insufficient Progress 9 1 20 may be terminated by [R&D] if [Client] hasnot selected a [Target] for further evalua-tion prior to the expiration of the ResearchPeriod

If Change in Key Employees 0 0 8 In addition, [Client] shall be entitled to ter-minate the Program upon 90 days’ noticeafter the date that any of [person], [person]or [person] is no longer obligated or able tocontinue to provide the same level of ser-vices as contemplated at the signing of thisAgreement.

Both/

Reversion Provision Orig. R&D Client

Reverts to Non-Breaching 12 0 0

Reverts to Non-Terminating 6 0 0

Failures Revert to 0 2 0

Ownership Reverts to 14 75 6

Except for: Breach 18 0

Change in Control 3 0

Bankruptcy 12 0

31

Tab

le5

Cor

rela

tion

sin

Con

trac

tC

hara

cter

isti

csT

his

table

pre

sents

pair

wis

eco

rrel

ations

bet

wee

nco

ntr

act

ing

vari

able

s.C

orr

elati

ons

den

ote

dw

ith

stars

are

signifi

cant

at

the

10%

level

.(D

iagonalel

emen

tsare

om

itte

d.)

Pro

ject

stage

isa

vari

able

that

equals

1if

the

pro

ject

isdis

cover

yst

age

(107

dea

ls),

2if

itis

lead

mole

cule

stage

(11),

3if

pre

-clinic

al

(7),

or

4if

phase

IFD

Atr

ials

(2dea

ls).

Pro

ject

Ter

mis

the

expec

ted

length

of

the

allia

nce

as

state

din

the

contr

act

at

signin

g.P

roje

ctsi

zeis

the

tota

lam

ount

of

financi

aloutlays

(both

act

ualand

conti

ngen

t)th

at

the

clie

nt

ple

dges

toth

eR

&D

.E

quity

0/1

isan

equity

dum

my,

while

Equity

stake

mea

sure

sth

esi

zeofth

eeq

uity

purc

hase

inm

illions.

Equity

fract

ion

mea

sure

sth

eeq

uity

valu

eas

afr

act

ion

ofth

evalu

eofth

eR

&D

firm

as

ofth

ela

sttim

eit

rece

ived

fundin

g.E

mplo

ym

ent

vari

able

sre

cord

whet

her

the

contr

act

stip

ula

tes

ace

rtain

num

ber

offu

ll-t

ime

equiv

ale

nts

(FT

Es)

,st

ipula

tes

thei

rle

vel

oftr

ain

ing

(PhD

s),or

nam

essp

ecifi

cin

div

iduals

as

pro

ject

mem

ber

s(N

am

es).

R&

DP

uts

refe

rsto

whet

her

the

equity

contr

act

allow

sth

eR

&D

toput

additio

nalsh

are

sto

the

clie

nt;

Warr

ants

refe

rsto

whet

her

the

clie

nt

has

the

right

toin

crea

seits

equity

stake.

Thes

eare

only

defi

ned

for

equity

dea

ls.

Pro

ject

Equ

ity

R&

DV

CFir

mR

&D

War

-R

oy-

Up-

Em

ploy

men

t

Stag

eTer

mSi

ze0/

1St

ake

Frac

.A

geFr

ac.

Val

ueP

uts

rant

sal

ty?

fron

t?FT

Es

PhD

s

Pro

j.Ter

m-0

.16*

Size

0.08

0.55

*

Equ

ity:

0/1

-0.1

30.

25*

0.23

*

Stak

e-0

.04

0.50

*0.

48*

0.56

*

Frac

.-0

.14

0.40

*0.

22*

0.53

*0.

48*

R&

DA

ge0.

17*

0.02

0.08

-0.2

1*0.

03-0

.14

VC

Frac

.-0

.06

-0.1

2-0

.20*

-0.0

7-0

.17

-0.0

1-0

.54*

Fir

mV

alue

0.15

0.29

*0.

41*

-0.0

10.

34*

-0.2

4*0.

18*

-0.3

6*

R&

DP

uts

-0.1

20.

40*

0.36

*.

0.31

*0.

23*

0.06

-0.0

50.

06

War

rant

s-0

.11

0.29

*0.

17.

0.26

*-0

.05

-0.0

5-0

.10

0.50

*0.

17

Roy

alty

?0.

47*

0.09

0.21

-0.1

30.

320.

170.

49*

-0.2

50.

41*

0.42

0.21

Upf

ront

?0.

03-0

.17*

-0.1

60.

140.

02-0

.07

-0.0

2-0

.01

-0.1

10.

09-0

.03

-0.1

4

FT

E?

-0.0

7-0

.04

-0.0

10.

070.

01-0

.03

0.08

0.05

-0.1

4-0

.02

0.06

0.02

-0.0

3

PhD

s0.

090.

010.

020.

070.

010.

07-0

.08

0.10

-0.1

6-0

.12

0.03

-0.2

4-0

.02

0.09

Nam

es0.

000.

140.

150.

130.

19*

0.23

*-0

.11

-0.0

3-0

.11

-0.0

50.

11-0

.01

-0.0

60.

18*

0.69

*

32

Table 6Determinants of Contract Length

The dependent variable measures the length of the contract analysis in bytes. Meetingfrequency records the number of times per year that the oversight committees set up tooversee the research projects meet. Board seat allocated is a dummy variable for whetheran equity stake includes a board seat. Equity is a dummy for equity participation, whileequity stake measures the dollar value of the equity stake, in millions. Firm value isthe value of the R&D firm as recorded by the Recombinant Capital funding database.Publicly traded is a dummy for whether the R&D is publicly traded. Upfront dummyand amount record whether an upfront payment took place, and if so, how large. Robustt-statistics are reported in parentheses.

Length of Contract

(1) (2) (3) (4)

Frequency 0.075 0.928 1.490 1.335

(0.25) (2.52)* (2.15)* (2.09)*

Board Dummy -1.650 -0.568 2.199 2.182

(0.60) (0.20) (0.80) (0.82)

Equity -1.690 -1.227 -3.501 -3.088

(0.79) (0.55) (1.45) (1.35)

Equity Stake 0.659 0.568 0.453 0.380

(3.03)** (3.22)** (3.06)** (2.84)**

Firm Value 0.019 0.026 0.025

(1.77) (2.00) (2.20)*

Publicly Traded -0.341 -1.329 0.020

(0.15) (0.52) (0.01)

Upfront Dummy 0.616 -0.783

(0.28) (0.36)

Upfront Amount 0.712 1.369

(2.28)* (3.74)**

Project Stage -4.517

(2.63)*

Constant 18.526 13.034 11.204 16.839

(10.78)** (5.99)** (3.95)** (4.82)**

Observations 99 72 55 55

R-squared 0.19 0.37 0.49 0.54

Robust t statistics in parentheses

* significant at 5%; ** significant at 1%33

Table 7Determinants of Upfront Payments

The dependent variable is the amount of the upfront payment from the pharmaceuticalto the biotech in millions of US dollars. Project stage takes on the values 1, 2, 3, or 4according to whether the project is discovery stage, lead molecule, pre-clinical, or in phaseI FDA trials (see table 5.) Publicly traded is a dummy for whether the biotech firm ispost-IPO, and firm age is the age of the biotech. Equity dummy and stake, respectively,measure whether the pharmaceutical took an equity position in the biotech, and thesize in millions of US dollars. Biotech’s Centrality measures the number of connectionsthe biotech firm has to other firms in the industry. Following Bonacich (1987), , for anNtimesN binary matrix X in which xij = 1 if an alliance has occurred between firm iand j, we define target centrality as:

Centrality = c(α, β) = α(I− βXt)−1Xt1.

where α is a scaling factor, 1 is a column vector of ones, and β is the weight placed onmore distant ties. The variable α is determined by solving the following equation:

α =

√n∑

c2(1, β)

where n is the number of firms in the network. See Robinson and Stuart (1999) fordetails.

Upfront Payment, $ millions

(1) (2) (3) (4)

Project Stage 2.192 3.261 3.366 3.336

(2.84)** (3.95)** (4.22)** (4.09)**

Publicly Traded 0.194 -2.330 -2.983 -2.924

(0.19) (1.76) (2.26)* (2.20)*

Firm Age -0.000 -0.000 -0.000 -0.000

(0.50) (0.61) (0.67) (0.74)

Biotech Centrality 2.549 2.852 2.768

(2.73)** (3.12)** (2.96)**

Equity Dummy -2.353 -2.409

(2.50)* (2.13)*

Equity Stake 0.023

(0.28)

Constant -2.656 -4.431 -2.901 -2.794

(2.37)* (3.51)** (2.20)* (2.07)*

Observations 98 90 90 87

Absolute value of t statistics in parentheses

* significant at 5%; ** significant at 1%

34